基于高光谱遥感的农作物分类研究进展

Research advances on crop identification using hyperspectral remote sensing

【目的】农作物类型识别是农作物面积、长势监测与产量预测的重要前提。及时、准确地获取农作物类型、空间分布以及种植面积对制定农业政策、促进社会经济发展和保障国家粮食安全具有重要意义。近年来,高光谱遥感凭借光谱分辨率高、光谱信息丰富等优点,已广泛应用于农作物制图中。【方法】文章归纳了高光谱遥感应用于农作物分类的研究进展,总结了国内外农作物分类常用的高光谱数据源,并分析了各种数据源的适用范围。梳理了农作物高光谱遥感分类方法,讨论了各种分类方法的优缺点。【结果】现有农作物高光谱遥感分类研究存在一些不足:(1)机载高光谱影像光谱分辨率高,但影像监测面积小,不适合大区域农作物面积提取研究;(2)星载高光谱影像监测面积较大,但空间分辨率较低,某些农作物面积提取实际应用中精度较低;(3)由于缺乏对农作物高光谱特征的研究,导致分类算法机理性不足,普适性较差。【结论】农作物高光谱遥感分类未来研究方向是:(1)丰富高光谱遥感监测的农作物类型;(2)提高高光谱影像的空间分辨率,实现农作物种植结构复杂、地块破碎地区的农作物分类研究;(3)进一步研究利用高光谱遥感进行农作物分类的机理和多源数据融合的方法。

[Purpose]Crop type identification is an important prerequisite for crop area,growth monitoring and yield forecasting.Timely and accurate access to crop types,spatial distribution and acreage is important for developing agricultural policies,promoting social and economic development,and ensuring national food security.In recent years,hyperspectral remote sensing has been widely used in crop mapping due to its high spectral resolution and rich spectral information.[Method]This paper reviews the research progress of hyperspectral remote sensing applied to crop classification,summarizes the hyperspectral data sources commonly used in crop classification at home and abroad,and analyzes the applicable range of various data sources.The method of hyperspectral remote sensing classification of crops was sorted out,and the advantages and disadvantages of various classification methods were discussed.[Result]There are some shortcomings in the classification of existing crop hyperspectral remote sensing.(1)The spectral resolution of airborne hyperspectral imagery is high,but the image monitoring area is small,which is not suitable for large area crop area extraction research.(2)The spaceborne hyperspectral imagery has a large monitoring area,but the spatial resolution is low,and the precision of some crop area extraction is lower in practical applications.(3)Due to the lack of research on the hyperspectral characteristics of crops,the classification algorithm is not systematic and the universality is poor.[Conclusion]Future research direction of crop hyperspectral remote sensing classification.(1)Enriching crop types monitored by Hyperspectral Remote Sensing.(2)Improving the spatial resolution of hyperspectral images,and realize the classification of crops with complex crop planting structures and broken areas.(3)Further research on the mechanism of crop classification and multi-source data fusion using hyperspectral remote sensing.